[2250] in Kerberos_V5_Development
Re: Function pointers with krb5_{calculate,verify}_checksum
daemon@ATHENA.MIT.EDU (Ken Raeburn)
Wed Feb 19 22:57:17 1997
To: Ken Hornstein <kenh@cmf.nrl.navy.mil>
Cc: krbdev@MIT.EDU
From: Ken Raeburn <raeburn@cygnus.com>
Date: 19 Feb 1997 17:12:52 -0500
In-Reply-To: Ken Hornstein's message of Wed, 19 Feb 1997 09:50:49 -0500
Ken Hornstein <kenh@cmf.nrl.navy.mil> writes:
> However, when trying to get Kerberos working on our HP/Convex Exemplar, I
> ran across what appears to be a problem with the way the checksum macros
> are declared. I kept getting bus errors anytime checksums were calculated.
...
> So this ends up being:
>
> *krb5_cksumarray[ctype]->sum_verf_func ?
> *(*krb5_cksumarray[ctype]->sum_verf_func)(...) : (abort(), 1);
>
> Now, color me stupid, but that sure _looks_ like one too many dereferences
> to me.
ANSI C rules specify that function "values" are automatically
converted to the function's address, except in the function-call
context. And a call done with a function pointer does the right
thing. I'm a bit fuzzy on those parts of the rules, but I think
that's roughly right.
So "*f", where f is a function pointer, has the dereference
automatically undone, sort of. Which means you get a function pointer
again. Repeat as needed. This code has two perfectly normal (if
somewhat odd-looking) function calls:
extern int (*f) ();
void x () { (***********f) (); f (); }
> Removing the * fixes the problem and makes Kerberos clients work.
>
> Am I way off base here? The thing that puzzles me is, if this is a problem,
> how come it doesn't show up on other architectures?
Because the other compilers implement this correctly.
